Re: ip_no_pmtu_disc and UDP

the UDP packet size is about 768, here is how packet path  like:

client <----------------------------------------router<-------------------------------------------------->server
(eth0 mtu 1500 ip 10.3.72.69)     (eth0 mtu 1500 ip 10.3.72.1,
          (eth0 mtu 1500 ip 10.2.72.99)
                                                      eth1.1102 mtu
567 ip 10.2.72.139)


UDP client test script:

#!/usr/bin/perl

use strict;
use warnings;
use IO::Socket::INET;

my $socket = IO::Socket::INET->new(
                                      PeerPort  => 9999,
                                      PeerAddr  => '10.2.72.99',
                                      Proto     => 'udp',
                                      )
                                  or die "Can't bind : $@\n";



$| = 1;

my $data = "012345670123456701234567012345670123456701234567012345670123456701234567012345670123456701234567012345670123456701234567012345670123456701234567012345670123456701234567012345670123456701234567012345670123456701234567012345670123456701234567012345670123456701234567012345670123456701234567012345670123456701234567012345670123456701234567012345670123456701234567012345670123456701234567012345670123456701234567012345670123456701234567012345670123456701234567012345670123456701234567012345670123456701234567012345670123456701234567012345670123456701234567012345670123456701234567012345670123456701234567012345670123456701234567012345670123456701234567012345670123456701234567012345670123456701234567012345670123456701234567012345670123456701234567012345670123456701234567";

$socket->send($data);


sleep(10);

$socket->close();

so I am hoping if I echo 0, 1, 2, 3 respectively to
/proc/sys/net/ipv4/ip_no_pmtu_disc, I am expected to see DF bit
set/unset from the client and should have shown me on the router eth0
interface tcpdump, but instead, DF bit never set on the client. am I
misunderstanding something?


for example:

 two concurrent tcpdump on router eth0 (mtu 1500) and eth1.1102 (mtu
576) interface:

1 #tcpdump -nn -i eth0 -v udp and host 10.3.72.69 &

14:51:11.946143 IP (tos 0x0, ttl 64, id 7193, offset 0, flags [none],
proto UDP (17), length 796)
    10.3.72.69.43748 > 10.2.72.99.9999: UDP, length 768


2# tcpdump -nn -i eth1.1102 -v udp and host 10.3.72.69 &

14:51:11.946164 IP (tos 0x0, ttl 63, id 7193, offset 0, flags [+],
proto UDP (17), length 572)
14:51:11.946176 IP (tos 0x0, ttl 63, id 7193, offset 552, flags
[none], proto UDP (17), length 244)
    10.3.72.69.43748 > 10.2.72.99.9999: UDP, length 768
    10.3.72.69 > 10.2.72.99: udp

as you can see, the router was fragmenting the UDP packet and not
sending icmp frag needed message, one reason I can think of is  the DF
bit is not set on the original UDP packet.

client is on kernel 4.3.0-rc7+, router is on kernel  3.13.0-rc3

On Fri, Oct 23, 2015 at 3:34 PM, Hannes Frederic Sowa
[off-list ref] wrote:

Hello,

On Fri, Oct 23, 2015, at 18:45, Vincent Li wrote:

quoted

It looks ip_no_pmtu_disc setting does not affect UDP IP packet DF bit
setting, is that intended behavior? echo 0, 1, 2, 3 respectively to
ip_no_pmtu_disc, UDP IP packet always have DF bit cleared, unless use
IP_PMTUDISC_DO on IP_MTU_DISCOVER as ip man page says.

Which size do the UDP packets have and what is your MTU? inet_create
also creates udp sockets and thus the setting does have effect.

quoted

in inet_create, seems to prove that.

       if (net->ipv4.sysctl_ip_no_pmtu_disc)
                inet->pmtudisc = IP_PMTUDISC_DONT;
        else
                inet->pmtudisc = IP_PMTUDISC_WANT;

so I am wondering why UDP is excluded by ip_no_pmtu_disc, why in
inet_create, not assign each individual ip_no_pmtu_disc setting to
inet->pmtudisc but only check true and assign IP_PMTUDISC_DONT or
IP_PMTUDISC_WANT only.

ip_no_pmtu_disc sysctl != IP_MTU_DISCOVER setsockopt. Also we cannot
change this as it would disrupt communication easily relying on this
established behavior.

See Documentation/ip-sysctl.txt:

ip_no_pmtu_disc - INTEGER
        Disable Path MTU Discovery. If enabled in mode 1 and a
        fragmentation-required ICMP is received, the PMTU to this
        destination will be set to min_pmtu (see below). You will need
        to raise min_pmtu to the smallest interface MTU on your system
        manually if you want to avoid locally generated fragments.

        In mode 2 incoming Path MTU Discovery messages will be
        discarded. Outgoing frames are handled the same as in mode 1,
        implicitly setting IP_PMTUDISC_DONT on every created socket.

        Mode 3 is a hardend pmtu discover mode. The kernel will only
        accept fragmentation-needed errors if the underlying protocol
        can verify them besides a plain socket lookup. Current
        protocols for which pmtu events will be honored are TCP, SCTP
        and DCCP as they verify e.g. the sequence number or the
        association. This mode should not be enabled globally but is
        only intended to secure e.g. name servers in namespaces where
        TCP path mtu must still work but path MTU information of other
        protocols should be discarded. If enabled globally this mode
        could break other protocols.

        Possible values: 0-3
        Default: FALSE

Bye,
Hannes